Tuesday March 5 2013
Scuola Normale Superiore
Dipartimento di Informatica – Università di Pisa
Rouen Business School
We present an algorithm to estimate the parameters of multivariate ARMA, GARCH and stochastic volatility models. The approach is based on a moment estimator; a similar approach has already been suggested in literature for univariate GARCH but its generalization to multivariate models requires some more linear algebra machinery, especially in the field of matrix equations.
The resulting estimator is extremely fast to compute, in comparison to maximum-likelihood approaches. We also discuss methods to regularize and improve this estimate.